Skip navigation
Reason Machines has launched Reason, its AI coding agent designed to maximize work completed per…
0:00 / 0:36

Reason Machines has launched Reason, its AI coding agent designed to maximize work completed per…

🚨 AI News | TestingCatalog

@testingcatalog85K followers

Visit on X
251 likesDownloadOriginal post

17,105 views

Reason Machines has launched Reason, its AI coding agent designed to maximize work completed per dollar. During testing, Reason solved more tasks than Codex on the same model while spending 22% less. Users can give it standing jobs to run autonomously, such as reviewing every new pull request, and run it using their own ChatGPT, Claude, GitHub Copilot, or Grok membership.

32 Comments

Comments from the original post on X (Twitter)

  • 🚨 AI News | TestingCatalog's profile picture

    @testingcatalog

    The Pro plan costs $50 a month and starts with a 3-day free trial. Test it out 👀

    8 likes

  • Luca Iaconelli's profile picture

    @luca_iaconelli

    No runaway bills on a membership at least. On my $200 Codex plan OpenAI keeps resetting the limits lately, so I almost always have usage left 😅

    2 likes

  • delulaᴗ✧'s profile picture

    @0xdelula

    better task efficiency with lower costs is worth exploring

  • Callahan's profile picture

    @kokyiup

    multi-step agents still struggle with error recovery on long tool chains, though latency is improving fast

  • Muhammed Kandeger's profile picture

    @Mkandeger2

    Work per dollar is the benchmark people actually feel. 22% cheaper on the same model means the harness matters as much as the model now. Would you let it review every PR unattended?

  • SAJJAD's profile picture

    @sajbrz

    Parallel tool execution cuts token waste and latency significantly

  • Last Son's profile picture

    @__Vik_tor

    Cost efficiency deserves more attention.

  • Naly's profile picture

    @NalyMetaX

    Hooking up my ChatGPT sub, easy win

  • Mihai Daniel Marius's profile picture

    @MIhaiDanielWeb3

    More tasks solved than Codex on the same model while spending 22% less, per their testing. Cost per finished task is the metric that matters now, not benchmark screenshots.

  • E_genius's profile picture

    @yevloop

    same model and 22% cheaper, so codex was just burning my tokens on vibes

  • Tekon Burns's profile picture

    @TekonBurns

    22% less spend for more solved tasks is huge

  • Moriarty's profile picture

    @Moriarty_tech

    $4.89 per attempt is API pricing. On a Codex sub nobody sees that bill, so this is for teams on raw API

  • bokensha's profile picture

    @iambokensha

    Work per dollar is what matters to me. Hitting a usage limit in the middle of a build kills the whole day

  • foundrceo's profile picture

    @foundrceo

    22 percent cheaper and more tasks solved is the stat people actually switch for

  • Della's profile picture

    @Della0x

    This could make coding much more efficient

  • Vincenzo Nudo's profile picture

    @vincenzo_naked

    A standing job on every PR is where the bill grows quietly, since it runs on the one line fixes too. I'd want a monthly cap on that job before caring about 22% per task.

  • Ꮧ Ꮦ ᎷᎧᏕᏖᏗᏦᎥᎷ's profile picture

    @mostakim93213

    curious to see how this impacts cost efficiency in the long run

  • Artzy's profile picture

    @heyartzy

    Using existing memberships makes trying it out painless

  • kayla ⛧⃝'s profile picture

    @KaylashowDq

    Less spending really matters

  • 𝗦𝘆𝗶𝗳's profile picture

    @syIFeth

    Less spend per solved task is the real flex